CEA & partners present ‘powerful step towards industrialisation’ of linear Si quantum dot arrays

In an invited paper, “Specificities of FDSOI QD Arrays Integration and Characterization”, presented at the 2022 IEEE VLSI Symposium on Technology & Circuits, researchers from CEA-Leti, Université Grenoble Alpes,CNRS Institut Néel, and CEA-Irig shared a new three-step characterisation chain for linear silicon quantum dot (QD) arrays fabricated on fully depleted silicon-on-insulator (FDSOI) material.

The team also offered several proposals for methodologies and metrics that can accelerate learning cycles at 300K (equivalent to 80 degrees F or 27 degrees C), while generating statistical data on expected device performance at cryogenic temperatures.

The three-step approach maximises the effectiveness of higher-temperature testing as part of the broader push towards making silicon QD devices manufacturable, by enabling developers to detect and analyze issues at the earliest and simplest point. “It’s a powerful step towards industrialization,” noted Dr. Maud Vinet, Quantum Computing Programme Director at CEA-Leti.

The initial wafer-level room temperature characterization step, which uses transistor-like testing protocols to gather data in a matter of hours, is followed by a more time-consuming wafer-level QD characterisation step at less than 2 degrees K, and a die-level qubit manipulation step (which can take days per device) at under 100mK.

The research team used the process to assess several considerations related to production of integrated QD arrays and make recommendations for addressing them.

One such consideration relates to the proposed use of linear arrays of floating-gate QDs to operate in a manner similar to a single-gate standard transistor. The researchers found that inner gates in these arrays offer consistent performance on threshold voltage (V_th) and subthreshold slope (SS), but outer gates exhibit more variability. The paper proposes dealing with these edge effect peculiarities (which can be caused by factors like random dopant fluctuation) by using the outer gates as access gates rather than for confinement of QDs.

Additionally, the paper notes that while the split-gate design being explored for linear QD arrays offers several functional advantages, its successful implementation will require very strict overlay control on one specific lithography step to achieve good symmetry, which is needed for consistent performance.

A third recommendation is focused on the issue of spurious dots within the qubit layer — a major source of yield loss in silicon QD arrays. Spurious dots can be detected during cryogenic testing, but revealing the inter-gate defectivity that causes them earlier in the characterisation chain (such as during 300K tests) would greatly accelerate the learning cycle. Although standard transistor parametric tests are unsuitable for the task, the researchers developed a 300K voltage-sweep technique capable of monitoring the screening effect inter-gate defectivity has on exchange gate polarization.

One key strength of the FDSOI material used in the group’s efforts, said Vinet, is the fact that back gates can be used to draw charges away from the interfaces. Back gates are typically fabricated using dopant implantation, which has the potential to introduce defects or parasitic dopants in the qubit layer. An alternative fabrication approach, using a TSV-like metallic back gate electrode, would be a way to alleviate this drawback while also enabling back-biasing, she added.

“These research results represent a significant step towards addressing the broader silicon spin qubit integration challenges we discussed at last December’s IEDM conference,” said Heimanu Niebojewski, CEA-Leti Lead Device Engineer. “It’s a very encouraging sign of the technology’s maturation.”